Tweaked version of Yace??? Thanks but no thanks.

[Grim Reaper]

Geschrieben von: / Posted by: Grim Reaper at 27 July 2000 19:23:32:
I've just downloaded both versions of Yace 22,
Thanks Dieter for a superb program, which just keeps getting better
however, running bench.bat with both programs its astonishing to see the
"tweaked" version by dann corbit is _10%_ slower on my AMD K6-2 500.
the same also happens on a P200 MMX.
We all appreciate your conversions of unix programs such as Amy, but what
are we supposed to do with these "tweaked" versions?
I would be interested to see other comparisons, maybe different hardware produce
a different story.

Danns compile
*************
6 tested, 4 found, 0 mates
29291693 nodes, 8048093 leavenodes in 176.08 sec 166352 n/s
test nodes 12064814 win nodes 2945673 mate nodes 0
win time 21.45 mate time 0.00 av depth 10.000 (nm 10.000) maxdepth 31, tu 849
log(win_nodes) = 16.47519 log(win_time) = -2.14164
Dieters compile
***************
6 tested, 4 found, 0 mates
32351656 nodes, 8755993 leavenodes in 176.11 sec 183702 n/s
test nodes 12849404 win nodes 2945673 mate nodes 0
win time 19.37 mate time 0.00 av depth 10.333 (nm 10.333) maxdepth 31, tu 856
log(win_nodes) = 16.47519 log(win_time) = -2.31952

[Grim Reaper]

Geschrieben von: / Posted by: Grim Reaper at 27 July 2000 19:41:46:
Als Antwort auf: / As an answer to: Tweaked version of Yace??? Thanks but no thanks. geschrieben von: / posted by: Grim Reaper at 27 July 2000 19:23:32:

I've just downloaded both versions of Yace 22,
Thanks Dieter for a superb program, which just keeps getting better
however, running bench.bat with both programs its astonishing to see the
"tweaked" version by dann corbit is _10%_ slower on my AMD K6-2 500.
the same also happens on a P200 MMX.
We all appreciate your conversions of unix programs such as Amy, but what
are we supposed to do with these "tweaked" versions?
I would be interested to see other comparisons, maybe different hardware produce
a different story.

Danns compile
*************
6 tested, 4 found, 0 mates
29291693 nodes, 8048093 leavenodes in 176.08 sec 166352 n/s
test nodes 12064814 win nodes 2945673 mate nodes 0
win time 21.45 mate time 0.00 av depth 10.000 (nm 10.000) maxdepth 31, tu 849
log(win_nodes) = 16.47519 log(win_time) = -2.14164
Dieters compile
***************
6 tested, 4 found, 0 mates
32351656 nodes, 8755993 leavenodes in 176.11 sec 183702 n/s
test nodes 12849404 win nodes 2945673 mate nodes 0
win time 19.37 mate time 0.00 av depth 10.333 (nm 10.333) maxdepth 31, tu 856
log(win_nodes) = 16.47519 log(win_time) = -2.31952

I did'nt read the message below, I had been warned - however I would still
be interested in bench speed comparisons with different hardware - anyone
tried this using Duron processor?
sorry Dann.

[Mogens Larsen]

Geschrieben von: / Posted by: Mogens Larsen at 27 July 2000 19:48:27:
Als Antwort auf: / As an answer to: Re: Tweaked version of Yace??? Thanks but no thanks. geschrieben von: / posted by: Grim Reaper at 27 July 2000 19:41:46:

I did'nt read the message below, I had been warned - however I would still
be interested in bench speed comparisons with different hardware - anyone
tried this using Duron processor?
sorry Dann.

On my Athlon, Dann's version is a couple of Knps faster. The difference is within the margin of error, so the speed is approx. the same. The best test would be PII and PIII processors, because the Corbit version is aimed at that architecture in particular.
Best wishes...
Mogens

[Volker Pittlik]

Geschrieben von: / Posted by: Volker Pittlik at 27 July 2000 20:53:41:
Als Antwort auf: / As an answer to: Re: Tweaked version of Yace??? Thanks but no thanks. geschrieben von: / posted by: Mogens Larsen at 27 July 2000 19:48:27:
...

On my Athlon, Dann's version is a couple of Knps faster. The difference is within the margin of error, so the speed is approx. the same.
The best test would be PII and PIII processors, because the Corbit version is aimed at that architecture in particular.
Best wishes...
Mogens

Same here on Celeron 520:
Dieter
------
6 tested, 4 found, 0 mates
40606334 nodes, 10990130 leavenodes in 175.26 sec 231697 n/s
test nodes 15245760 win nodes 2985044 mate nodes 0
win time 15.19 mate time 0.00 av depth 10.500 (nm 10.500) maxdepth 31, tu 866
log(win_nodes) = 16.48109 log(win_time) = -3.92873
Dann
----
6 tested, 4 found, 0 mates
40700351 nodes, 11041497 leavenodes in 175.31 sec 232166 n/s
test nodes 15341366 win nodes 2985044 mate nodes 0
win time 15.11 mate time 0.00 av depth 10.500 (nm 10.500) maxdepth 31, tu 866
log(win_nodes) = 16.48109 log(win_time) = -3.82551
Volker

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 27 July 2000 21:27:22:
Als Antwort auf: / As an answer to: Tweaked version of Yace??? Thanks but no thanks. geschrieben von: / posted by: Grim Reaper at 27 July 2000 19:23:32:

I've just downloaded both versions of Yace 22,
Thanks Dieter for a superb program, which just keeps getting better
however, running bench.bat with both programs its astonishing to see the
"tweaked" version by dann corbit is _10%_ slower on my AMD K6-2 500.
the same also happens on a P200 MMX.
We all appreciate your conversions of unix programs such as Amy, but what
are we supposed to do with these "tweaked" versions?
I would be interested to see other comparisons, maybe different hardware produce
a different story.

Danns compile
*************
6 tested, 4 found, 0 mates
29291693 nodes, 8048093 leavenodes in 176.08 sec 166352 n/s
test nodes 12064814 win nodes 2945673 mate nodes 0
win time 21.45 mate time 0.00 av depth 10.000 (nm 10.000) maxdepth 31, tu 849
log(win_nodes) = 16.47519 log(win_time) = -2.14164
Dieters compile
***************
6 tested, 4 found, 0 mates
32351656 nodes, 8755993 leavenodes in 176.11 sec 183702 n/s
test nodes 12849404 win nodes 2945673 mate nodes 0
win time 19.37 mate time 0.00 av depth 10.333 (nm 10.333) maxdepth 31, tu 856
log(win_nodes) = 16.47519 log(win_time) = -2.31952

In my message, I said it requires an Athlon if you use AMD.
Sometimes my versions are faster, and sometimes slower. It does really well on a Xeon. But you should always test them. You will find (for instance) that my ExChess 3.14 is much faster than the one from the author's home page. So (as always) your mileage may vary.
I suggest that if you don't want to bother testing them, then just always use the version provided by the original author. You really can't go wrong that way.
I make other changes too. Sometimes, I make changes to quiet Lint's groanings that affect performance and may not really make the program any more robust (they might even introduce errors).
Sometimes, my page is the only place to find some particular binary.
I offer no support whatsoever to anything I provide.
Caveat Emptor.
But the price is right.


my ftp site